I want to build a 'simple' 3D game in Java for some language research that I'm doing. I know that the Java 3D community is in a state of flux with the slow fade of Java3D and the rise of jogl, in addition to constant evolution of API's/engines such as Xith3D, LWJGL, OpenMind, etc. I'm trying to determine the 'best' technologies to learn/pursue.
The rough idea for the 'game' is as follows:
1. create a virtual playroom - initially a single room is fine, but a small house would probably be more interesting
2. populate the room/house with some static and some movable objects (e.g. furniture & simple toys)
3. create a simple child character within the game representing the computer
4. create a simple parent character within the game representing the user
5. define some simple actions that the parent can perform on each object (e.g. pick up, drop, touch, bounce, etc.)
6. allow the user to describe textually to the computer what he/she is doing
The child/computer will 'watch' what is happening and attempt to learn basic language based on the current actions and corresponding descriptions of the parent/user.
All of this is based on a system I devloped in Java which enables a computer to learn nouns/verbs in a bottom-up fashion by 'watching' simple videos and parsing short descriptions of each video.
I believe that by moving to a virtual vision system, I can bypass a lot of computer vision issues and provide a much more robust language learning environment.
A short paper on my research is available from:http://www.cs.cornell.edu/~regina/lwm03/papers/lwm007.pdf
Any pointers on the 'best' approach (technologies, tools, etc.) for this project would be much appreciated.